The idea that non-organic entities are capable of independent thought and action isn’t a new concept. There are references as far back to the zombie-like golem of biblical times, and the modern version might be said to begin with the 1920 Karel Capek play “R.U.R.” (for “Rossum’s Universal Robots”), which introduced the term robot.
But perhaps the first thing many people think of when they hear the term artificial intelligence, or AI, is the self-aware computer HAL 9000 in the movie “2001: A Space Odyssey,” which refused to open the pod bay doors and allow astronaut Dave Bowman back into Discovery One. In reality, while we do have computers that sometimes do things that are unexpected, so far none have actively tried to kill someone.
Artificial what?
One big stumbling block when trying to specifically define AI is that the idea of what intelligence itself is, is rather nebulous. But a fairly commonly accepted definition of AI is a computer that is capable of making decisions on how to perform a new task based on what tasks have come before, and how they were resolved — in essence, a computer that can learn to perform tasks it may not have been specifically programmed to do. The part of AI where a computer can distill the actions currently being performed on a task is called machine learning. Applying this machine learning “knowledge” to tasks for which it may not have been specifically programmed is a vital part of artificial intelligence.
While AI today is far advanced over what it started out as, we’ve all seen some aspect of it in action over the years. Jessica Veiga, chief marketing officer at Botkeeper, pointed out, “First, acknowledge that you’re probably using AI in aspects of your everyday life already, and way more than you realize. Siri, Alexa, Google Assistant, your phone’s navigation, etc., are all examples of widely adopted AI. And while historically there’s always been a bit of a wariness of new technology in the accounting profession, let’s face it — AI is here, and it isn’t going anywhere.”
But examples of AI being used in computer applications go further back than this. The autocorrect function that drives you nuts in word processing and on your phone, and the auto-fill functionality that many accounting and bookkeeping applications have employed for years, are all examples of AI, albeit on an entry-level scale.
Understanding how AI has been implemented and will be implemented in accounting going forward requires a common understanding of what AI is, exactly — but that can be hard to find.
David Cieslak, a technology thought leader and chief cloud officer at RKL eSolutions, gave us a good start: “AI is a very broad term. It incorporates a number of subtopics, everything from machine learning to deep learning and natural language processing. I look at AI and think about it overall as just really a tool that analyzes data to ultimately help affect some kind of next action.”
Victoria Jones, AI evangelist at Zoho Corp., however, noted, “The thing is, everyone has a different definition of AI. Generally, when we talk about AI, we're talking about technologies that are able to take some kind of work or analysis off of the user's shoulders without having to be manually programmed. So, it's able to look at what somebody's doing, and adjust or offer up help based on that.”
Regardless of how AI is actually implemented, it requires the analysis of large amounts of data. The more data that’s available on how to perform an operation, analyze a trend, or calculate a financial position or the result of operations, the more accurately the AI application will be able to automate that process or offer help with it.
Cieslak noted that most current AI requires the application to be housed in the cloud. “AI really does its best work when there's not just a sufficient data set, but a large data set,” he explained. “So the more data you have and the more pattern recognition that takes place, the more it's going to be able to do the analysis and come up with something that seems to be a cohesive response.”
“AI transforms an accountant’s daily existence,” said Evan DeFord, director of expert services at Inflo. “Using AI as the workhorse allows accountants to focus on their highest and best use. By automating work and creating suggestions or even benchmarks, we are presenting users with all the salient data that they need to analyze, assess, or make decisions, but intentionally not replacing the professional judgment of the accountant.”
Let an artificial entity do it
Let’s face it — bookkeeping is necessary, but on the whole, pretty boring. But along with auditing, it’s also a primary area of application of AI in the accounting realm.
“As for accounting work specifically, there is an extraordinary use case for AI,” said Veiga. “One of the biggest issues we’ve seen our firm partners face is that ‘data dilemma.’ The bookkeeping and accounting work, which is of course the backbone to business decisions and planning, has historically been a huge headache. Collecting the needed data points, organizing them, and using them to create a picture of financial health has always required processes that are labor-intensive, error prone, and limiting to growth. There’s a really powerful use case for AI within the profession: streamlining repetitive tasks, reconciling transactions, organizing data, and others. As the role of the CPA moves further away from the number-crunching and more toward advisory services, accountants leveraging AI can get you out of the data entry, and into the data application.”
Cieslak agreed that AI is an appropriate technology to address this: “I look at [AI] and say I want to work smarter, better. How can I leverage that as a tool? “
Another increasingly popular accounting area where AI applications are being widely applied is in the field of auditing. Auditing has always been a resource-intensive application, with larger audits taking months or more to complete and requiring large teams of auditors and other personnel. Unlike much of accounting, auditing doesn’t consist of recording or forecasting financial transactions. Rather, it addresses the accuracy of how those transactions have been recorded, which is a very different animal.
“In audit and assurance-related work, data analytics and, more importantly, artificial intelligence applied to the risk assessment allows financial professionals to think beyond traditional sampling and legacy data analysis systems,” said Robin Grosset, chief technology officer at MindBridge. “Artificial intelligence allows auditors to automate risk discovery early on in their process to focus on insights based on pattern and anomaly detection of vast data sets. AI has the power to unfetter the data locked in the business and to be more impactful to the common financial assurance workflows. Continuous monitoring using such a technology can provide value at every stage of the audit process. Coupled with the work done today, accountants should be looking at the exposure drafts on changing standards which will impact the use of technology and make AI a critical tool in the auditors’ toolbox, so start now and be ready for those changes.”
DeFord added further justification for the use of AI: “Pre-emptively identifying client transactions as noteworthy allows the user to dictate whether they warrant further attention. Or being able to use AI to suggest more powerful visualizations to automate reports allows users to harness the power of AI at the appropriate point in their workflow. Allowing accountants to spend more time at that professional judgment or decision-making step in assessing the information, rather than spending more of the time gathering the information to make the assessment, is fundamentally important.”
One big advantage to using AI in the audit process is that it allows a deeper dive into the transactions that would either normally require a large application of labor, or possibly be missed because the discrepancy is below an amount level.
Samantha Bowling, a partner at Garbelman Winslow CPAs said, “I guess for me, the eye-opener was when I had this audit for a nonprofit, and their biggest fear or risk is credit card fraud where officers use credit cards for personal use and things like that. So, when I used this platform [MindBridge], it kept flagging these credit card transactions and they were way below my materiality level. Normally I wouldn't have even looked at them, to be honest. I mean, just by chance, if I selected that transaction or that credit card statement, maybe I would've seen it. My guess is not. It flagged it because, in the five-year history of the information I uploaded, they never charged to this vendor before, ever. That was the only reason it flagged it — it was something new that it didn't recognize and it was coded to an expense that was miscellaneous or office expense. Even with AI, you're not looking at everything, but the AI platform is analyzing everything based on risk. So, in a way you're doing a better risk-based audit because you have the whole population. Now I'm not going to lie — while it doesn't work for every general ledger or every accounting system, I feel like it's come a long way.”
AI and OCR — a lovely couple
One interesting point that emerged from the experts we spoke to was the importance of optical character recognition, or OCR, in AI. Given that the AI functions of the software need to be able to discern what you are doing and where it can assist, it first has to accurately “read” your input and actions. OCR today is more accurate than ever, which means that there are more actions that can be completed with better accuracy. And even if the AI in an application is not actively performing a helpful or time-consuming action, it can be used to capture an unexpected input or request, or make data entry faster, easier and more accurate.
“OCR, once it's trained up all the way, doesn't get tired after eight hours of processing data,” Zoho’s Jones noted. She also mentioned expense reporting and recording as one area where OCR and AI are particularly useful. “Zoho has built up in-house OCR technology that shows up in a lot of developer tools and inventory tools, but also in keeping track of expenses. And for expenses, that's one of the No. 1 places where you need to just take as many barriers out of the way as possible, because people just love to not do them.”
“We spent years and years of having to collect the data, take a look at what the OCR said, correct anything that was wrong, and then coming out the other side after this whole long process and being able to say, ‘All right, well this is now more accurate than someone sitting at their desk for eight hours and just getting tired and getting tired of looking at receipts the whole time,’” she continued. “And I think when it comes to financial applications, accuracy is more important than fast. It's one area where if it takes a little longer to be right, then you take the time to do it. And I think that's an area that AI can be focused on helping the accountant be more accurate. It's like having another pair of eyes on your work instead of having someone doing the work for you.”
Just scratching the surface
While we’ve touched on a few of the accounting-oriented applications where AI is being effectively used, there’s one broad area that we haven’t referenced, and that’s financial analysis and forecasting. AI is being heavily used in making sense of the numbers and helping accountants and financial analysts apply their expertise to improving business processes, operations, and profitability. But that, in itself, is worth at least a separate article (or two).
The experts we spoke with were very upbeat about the future of AI. Cieslak believes that 2022 could very well be a watershed year: “I look at AI in accounting, AI in general and say, ‘I think it's really one of the most exciting topics we have in front of us.’”
And if the present-day applications are any indication, Cieslak is probably not only right, but might be understating the effect that AI will have on the profession.
“I think that we’re going to see ‘predictive’ advisory become more of the norm as AI advances,” said Botkeeper’s Veiga. “Running a business will become less reactive, and far more proactive when it comes to decision-making. Realistically, there have been huge leaps in the sophistication of AI in just the last few years alone. I’d say we’re about five years away from AI really becoming the standard within most accounting firms for not only the repetitive tasks such as data entry, but also for client guidance and/or predictive advisory.”
Liz Mason, founder and CEO of High Rock Accounting, offered a legislative and regulatory caveat against too-high expectations: “I think that there's a lot of opportunity in the future, but I do think it's going to take longer than the technology is developing because the rate of technology and the rate of change is only escalating. So the tech is developing so much faster than we can even keep up with. When you see that curve and then you see the change adoption curve for accountants, the gap is just widening. When you think about the perspective of ‘What is this going to look like in 10 years?’ Well, in 10 years, hopefully we have guidance changes and hopefully we have legislative changes, but in reality, it's probably going to be very incremental compared to the change in technology. So [at] this moment there is, in my opinion, no way to keep up with the rate of change in tech with the guidance that we have.”
She continued, “You can't just take a piece of AI and implement it into the same process that you have on the audit side. You have to actually update how you're doing it and the way that you're thinking about it. In order to do that, we need to see change in legislation and then we need to see change in the way that the AICPA views it and what they're advising audit firms to do. All of that has to happen to actually effectively implement the AI and the technology that we have to make it work.”
Inflo’s DeFord also noted, "We need to have intelligence before we have artificial intelligence. And what I mean by that is, I don't care if you started your career 100 years ago or five years ago, some of this is really tough. And the new person on the job goes, ‘Well, how did you know to ask this, or to look at this, or to document it this way?’ And it's hard to answer. You go, ‘Well, it just takes time. You just kind of learn it.’”
Jones added a final observation: “One of the ways that I think helps to think about AI technologies is that everyone — the tools, the people using the tools, the people who have the tools being used — all want the same thing. And that is for everything to be organized without having to think about it. Thinking about AI tools as a way to just make it faster for auditors to be able to go through and do their job. And to make it faster for accountants to be able to square everything up and make sure that everything is in the right place. It's really just about making people's jobs easier so that they can spend more time doing the things that are hard for people, but not necessarily hard for machines.”
